FormalPara Key Summary Points
Why carry out this study?
To help manage blood glucose, people with type 2 diabetes have glucose-lowering medications added in a stepwise manner as their disease progresses; early use of combination therapy has recently been advocated in international guidelines.
Oral semaglutide is the first oral glucagon-like peptide-1 receptor agonist and is likely to be used in combination with other glucose-lowering agents.
In the current post-hoc subgroup analyses, data from the PIONEER clinical trial programme were used to explore the efficacy and safety of oral semaglutide versus comparators in patients receiving various combinations of background medications and background insulin regimens.
What was learned from the study?
Oral semaglutide had similar efficacy and tolerability regardless of the background medication in PIONEER 3–5, 7 and 8, and the insulin regimen in PIONEER 8. Oral semaglutide generally improved glycated haemoglobin and body weight parameters to a greater extent than comparator treatments across background medication and insulin regimen subgroups and trials.
Oral semaglutide appears suitable for a broad population of patients with type 2 diabetes in combination with other glucose-lowering agents.

Digital Features

This article is published with digital features, including a summary slide and graphical abstract, to facilitate understanding of the article. To view digital features for this article go to https://doi.org/10.6084/m9.figshare.13477029.

Introduction

Semaglutide is a glucagon-like peptide-1 (GLP-1) analogue that is the first and, thus far, only GLP-1 receptor agonist (GLP-1RA) to be formulated as a once-daily oral tablet for people with type 2 diabetes (T2D) [1,2,3,4]. GLP-1RAs are generally recommended for patients with T2D whose glycated haemoglobin (HbA1c) is above target despite diet, exercise and existing glucose-lowering medication, and GLP-1RAs with proven cardiovascular benefit are also recommended in patients with T2D at elevated cardiovascular risk, regardless of HbA1c [5, 6]. Consequently, patients receiving oral semaglutide will typically already be receiving treatment with other glucose-lowering agents. It is therefore important to understand how the efficacy and safety of oral semaglutide may vary based on the presence of concomitant treatments.

The efficacy and safety of oral semaglutide was investigated in the global phase 3a Peptide InnOvatioN for Early diabEtes tReatment (PIONEER) programme [2]. The PIONEER 3, 4, 5, 7 and 8 trials included patients with T2D who were using more than one background medication [7,8,9,10,11]. In PIONEER 3 and 7, oral semaglutide 7 mg, 14 mg, and flexibly dosed significantly reduced HbA1c and body weight compared with sitagliptin 100 mg (up to week 52 in PIONEER 7 and week 78 in PIONEER 3) [7, 10]. In PIONEER 4, oral semaglutide 14 mg was noninferior to liraglutide 1.8 mg and superior to placebo in decreasing HbA1c, and superior to both for reducing body weight, at week 26 [8]. In PIONEER 5 in patients with moderate renal impairment, and PIONEER 8 in patients receiving insulin, oral semaglutide was superior to placebo for reducing HbA1c and body weight at week 26 [9, 11]. The safety profile of oral semaglutide in these trials was similar to that expected of other GLP-1RAs, with tolerability issues mainly comprising mostly transient, mild-to-moderate gastrointestinal events [7,8,9,10,11].

The aim of the current exploratory analyses was to investigate whether different combinations of background medications and insulin regimens affected the efficacy and safety of oral semaglutide compared with active comparators and placebo in the PIONEER 3, 4, 5, 7 and 8 trials.

Methods

Trial Designs

The designs of the PIONEER 3–5, 7 and 8 trials have been reported previously and are summarised in Table 1 [7,8,9,10,11]. In brief, patients were adults who had been diagnosed with T2D ≥ 90 days before screening, had baseline HbA1c above target (7.0–10.5% [53–91 mmol/mol] in PIONEER 3 and 7.0–9.5% [53–80 mmol/mol] in PIONEER 4, 5, 7 and 8) and were receiving stable doses of glucose-lowering background medication(s). Patients in PIONEER 5 had moderate renal impairment (estimated glomerular filtration rate [eGFR] 30–59 ml/min/1.73 m2).

Table 1 Trial designs for PIONEER 3, 4, 5, 7 and 8

The PIONEER trial protocols were approved by institutional review boards or independent ethics committees at the participating trial sites, and the studies were conducted in accordance with ICH Good Clinical Practice guidelines and the Declaration of Helsinki. All participants provided written informed consent before undertaking trial-related activities.

In PIONEER 3, patients were randomised to oral semaglutide 3, 7 or 14 mg once daily or sitagliptin 100 mg once daily added to background metformin with or without a sulphonylurea [7]. In PIONEER 4, oral semaglutide 14 mg once daily was compared with both liraglutide 1.8 mg and placebo in patients receiving metformin with or without a sodium-glucose cotransporter-2 inhibitor (SGLT2i) [8]. PIONEER 5 was a placebo-controlled trial of oral semaglutide 14 mg once daily in patients with moderate renal impairment receiving metformin or sulphonylurea or both or basal insulin with or without metformin [9]. In PIONEER 7, flexibly dosed oral semaglutide 3, 7 and 14 mg once daily were compared with sitagliptin 100 mg once daily in patients receiving 1–2 of the following oral glucose-lowering medications: metformin, thiazolidinedione, sulphonylurea and SGLT2i [10]. Finally, in PIONEER 8, patients were randomised to oral semaglutide 3, 7 or 14 mg once daily or placebo as an add-on to basal, premixed or basal-bolus insulin with or without metformin [11].

Randomisation in these trials was stratified by baseline background medication. In the current post-hoc subgroup analyses, data from the PIONEER clinical trial programme were used to explore the efficacy and safety of oral semaglutide versus comparators in patients receiving various combinations of background medications and background insulin regimens. Data from PIONEER 1 were not included because oral semaglutide was given as monotherapy [12], PIONEER 2 was excluded because all patients randomised to oral semaglutide or comparator were receiving background metformin (and no other background glucose-lowering medication) [13], and PIONEER 6 was excluded because it was a cardiovascular outcomes trial [14].

Subgroup Analyses

All randomised patients in PIONEER 3–5, 7 and 8 were included in the main background medication subgroup analyses, which were done for each trial separately. Subgroup data were not pooled across trials because of substantial differences in study design and patient population. Randomisation was stratified according to patients’ background medications: metformin with or without sulphonylurea in PIONEER 3; metformin with or without SGLT2i in PIONEER 4; metformin alone, sulphonylurea with or without metformin, or basal insulin with or without metformin in PIONEER 5; metformin, SGLT2i or thiazolidinedione with or without sulphonylurea in PIONEER 7; and insulin with or without metformin in PIONEER 8.

In addition, a separate subgroup analysis was performed for PIONEER 8 to explore the effect of the insulin regimen on the efficacy and safety of oral semaglutide compared with placebo.

Efficacy endpoints were change from baseline in HbA1c and body weight (and, for the separate PIONEER 8 subgroup analysis, change from baseline in insulin dose) at the planned end of treatment (weeks 26, 52 or 78, according to the trial). Efficacy data are presented for oral semaglutide 14 mg or flexibly dosed, and for comparators. Safety outcomes were analysed by subgroup across all trials for all patients (including those receiving oral semaglutide 3 and 7 mg).

Statistical Methods for Study-Level Analyses

The PIONEER programme employed the use of two estimands [15]. The trial product estimand evaluates the treatment effect for all randomised patients under the assumption that all patients remained on the trial product for the entire planned duration of the trial and did not use rescue medication. The treatment policy estimand evaluates the treatment effect for all randomised patients, regardless of trial product discontinuation or the use of rescue medication.

For the main background medication subgroup analyses, the trial product estimand was used to reflect the effect of oral semaglutide versus each comparator without the confounding effect of rescue medication. For the efficacy outcomes (HbA1c and body weight), treatment differences were estimated by a mixed model for repeated measurements (MMRM) using data collected prior to premature trial product discontinuation or initiation of rescue medication from all randomised patients. A restricted maximum likelihood was used. The independent effects included in the model were treatment, region, strata (renal function for PIONEER 5 and insulin regimen for PIONEER 8), background medication and interaction between treatment and background medication as categorical fixed effects and baseline value as a covariate, all nested within visit, and using an unstructured residual covariance matrix.

The MMRM is a well-established method that accounts for the uncertainty pertaining to missing data. This analysis assumes that the missing data mechanism is missing at random.

In PIONEER 8, patients whose insulin dose was increased by ≥ 20% during treatment were considered to have required rescue medication. Therefore, for the insulin regimen subgroup analysis, the treatment policy estimand was used to reflect the effect of initiating treatment with oral semaglutide compared with initiating treatment with placebo, both potentially followed by either discontinuation of the trial product and/or initiation of rescue medication. Treatment differences by insulin regimen were estimated by a pattern mixture model using multiple imputation to handle missing data. All data collected at the planned end of treatment (week 52), irrespective of discontinuation of the trial product or initiation of rescue medication, were included in the statistical analysis. Missing data were imputed within groups defined by trial product and treatment status at week 26. Both the imputation and the analysis were based on analysis of covariance models with treatment, region, stratum (with or without metformin), insulin regimen and interaction between treatment and insulin regimen as categorical fixed effects and baseline value as a covariate. The results were combined using Rubin’s rule [16].

p values were calculated for the unadjusted two-sided test of the treatment by background medication subgroup interaction. Estimated treatment differences (ETDs) with 95% confidence intervals were calculated, but p values are not presented because these comparisons were exploratory in nature and not controlled for multiplicity.

Safety data were analysed descriptively by treatment group (oral semaglutide 3 mg, 7 mg, 14 mg, flexibly dosed, and comparators) and background medication/insulin regimen subgroup for all patients in the trials for the incidence of adverse events (AEs), including serious AEs, gastrointestinal AEs and genitourinary AEs, as well as events leading to trial product discontinuation and hypoglycaemic episodes.

Results

Patient Disposition and Baseline Characteristics

In total, 2836 patients receiving oral semaglutide 14 mg/flexibly dosed or comparators were included in the background medication subgroup analyses, and 365 patients receiving oral semaglutide 14 mg or placebo in PIONEER 8 were included in the insulin regimen subgroup analysis. As may be expected, mean diabetes duration tended to be longer in patients receiving a greater number of background medications, ranging from 7.2 to 10.6 years in patients on metformin monotherapy to 8.4 to 17.3 years in patients receiving more than one background medication. Otherwise, baseline characteristics were generally similar across subgroups within each trial (Table 2(i)). As expected, based on the design and the inclusion criteria, patients in PIONEER 5 tended to be older (mean age ≥ 70 years) and had reduced renal function (mean eGFR 46–50 ml/min/1.73 m2) compared with those in the other trials (55–62 years and ≥ 90 ml/min/1.73 m2, respectively, across other trials and subgroups). In PIONEER 8, compared with patients receiving basal or premixed insulin, patients receiving basal-bolus insulin appeared to have a greater duration of diabetes (13 and 14 years vs 16 years, respectively), greater body weight (85 and 78 kg vs 89 kg) and a higher total daily insulin dose (35 and 48 U vs 79 U) (Table 2(ii)). The proportion of patients who completed treatment with rescue medication varied across trials, but was consistently lower with oral semaglutide 7  mg, 14 mg, and flexibly dosed (0–26% across background medication subgroups) than comparator treatments (7–34%), whereas the proportion of patients who discontinued the trial product was often somewhat higher (12–25% vs 5–15% except when sitagliptin was added to multiple background medications in PIONEER 7 [50% in a very small subgroup of 8 patients]) (see Table S1 in the “Supplementary Information”). There was no clear pattern to these intercurrent events by background medication subgroup.

Table 2 Baseline characteristics for the background medication subgroup analysis (i) and the insulin regimen subgroup analysis (ii)

Efficacy Outcomes by Background Medication Subgroup

HbA1c and body weight reductions were generally greater with oral semaglutide 14 mg and flexibly dosed than with comparators, although glycaemic efficacy was similar to liraglutide 1.8 mg. HbA1c reductions with oral semaglutide 14 mg and flexibly dosed were broadly similar across background medication subgroups within each trial; change from baseline ranged from − 1.0 to − 1.5% at the end of treatment (Fig. 1(i)). In comparison, changes from baseline were − 0.3 to − 0.9% with sitagliptin 100 mg and − 0.9 to − 1.1% with liraglutide 1.8 mg. ETDs between oral semaglutide and comparators were generally consistent across background medication subgroups within each trial. Background medication only had a significant effect on the change in HbA1c in PIONEER 8, where the treatment difference between oral semaglutide and placebo was smaller in patients taking insulin and metformin compared with insulin alone (Fig. 1(ii)).

Fig. 1
figure 1

Change from baseline in HbA1c and estimated treatment differences for oral semaglutide 14 mg and flexibly dosed versus comparators by background medication (i and ii) and by insulin regimen (iii and iv). (i) Data are estimated changes from baseline (means with SEs in brackets) for the trial product estimand (on trial product without rescue medication). The p value is for the unadjusted two-sided test of the treatment by subgroup interaction. N number of patients contributing to the analysis, Other includes thiazolidinediones and other oral glucose-lowering medications. (ii) Data are for the trial product estimand (assumes patients remained on the trial product without rescue medication use). (iii) Data are estimated changes from baseline (means with SEs in brackets) for the treatment policy estimand (regardless of trial product discontinuation or rescue medication use in all randomised patients). N total number of patients in each subgroup (full analysis set). (iv) Data are for the treatment policy estimand. CI confidence interval, ETD estimated treatment difference, flex flexible dose adjustment, HbA1c glycated haemoglobin, ins insulin, met metformin, SE standard error of the mean, SGLT2i sodium-glucose cotransporter-2 inhibitor, SU sulphonylurea

Body weight reductions with oral semaglutide 14 mg and flexibly dosed varied slightly across background medication subgroups within each trial, but with no significant treatment interactions by background medication (Fig. 2(i)). Changes from baseline in body weight were greater with oral semaglutide (− 2.2 to − 5.0 kg) than comparators (0.5 to − 2.1 kg with sitagliptin and − 3.0 to − 3.7 kg with liraglutide) across background medication subgroups. ETDs between oral semaglutide and comparators were generally consistent across background medication subgroups within each trial (Fig. 2(ii)).

Fig. 2
figure 2

Change from baseline in body weight and estimated treatment differences for oral semaglutide 14 mg and flexibly dosed versus comparators by background medication (i and ii) and by insulin regimen (iii and iv). (i) Data are estimated changes from baseline (means with SEs in brackets) for the trial product estimand (on trial product without rescue medication). N number of patients contributing to the analysis, Other includes thiazolidinediones and other oral glucose-lowering medications. (ii) Data are for the trial product estimand (assumes patients remained on the trial product without rescue medication use). (iii) Data are estimated changes from baseline (means with SEs in brackets) for the treatment policy estimand (regardless of trial product discontinuation or rescue medication use in all randomised patients). N total number of patients in each subgroup (full analysis set). (iv) Data are for the treatment policy estimand. CI confidence interval, ETD estimated treatment difference, flex flexible dose adjustment, ins insulin, met metformin, SE standard error of the mean, SGLT2i sodium-glucose cotransporter-2 inhibitor, SU sulphonylurea

Efficacy Outcomes by Insulin Regimen

At week 52, reductions in HbA1c and body weight with oral semaglutide 14 mg were greater than those with placebo when added to insulin (Figs. 1(iii), 2(iii)). ETDs for changes from baseline in HbA1c and body weight were generally similar across all insulin regimen subgroups, and no significant treatment interactions by insulin regimen were found (Figs. 1(iv), 2(iv)). At week 52, the total daily insulin dose was decreased from baseline with oral semaglutide 14 mg for patients on basal, premixed and basal-bolus insulin, and were generally consistent regardless of background insulin regimen (see Fig. S1 in the “Supplementary Information”).

Safety Outcomes

The incidences of any on-treatment AEs, gastrointestinal AEs, serious AEs and AEs leading to trial product discontinuation by treatment and background medication subgroup are shown in Table 3. The overall incidence of AEs was similar between oral semaglutide and comparators (approximately 60–80% of patients in any subgroup had an AE) and, as expected, gastrointestinal AEs (primarily nausea, vomiting and diarrhoea) made up a substantial proportion of the overall events for oral semaglutide. However, the incidence of gastrointestinal AEs was generally similar regardless of the background medication subgroup. Serious AEs occurred in 3–17% of patients, and up to 17% of patients permanently discontinued their assigned medication across background medication subgroups. There was no discernible pattern in the incidence of serious AEs and AEs leading to trial product discontinuation by treatment group or background medication subgroup. In the insulin regimen analysis, the proportion of patients with any on-treatment AE was somewhat higher in the patients receiving oral semaglutide added to basal-bolus insulin (82–87%) than among those receiving oral semaglutide plus basal (65–82%) or premixed (69–80%) insulin, but no such trend was evident in the incidence of gastrointestinal or serious AEs, or in those leading to trial product discontinuation (see Table S2 in the “Supplementary Information”).

Table 3 On-treatment adverse events

There was generally a low incidence of severe and symptomatic blood glucose-confirmed hypoglycaemia, but the rate was higher in background medication subgroups containing sulphonylurea or insulin, regardless of the trial product received (see Table S3 in the “Supplementary Information”). As expected, there was a higher incidence of hypoglycaemia in PIONEER 8, in which all patients were receiving background insulin, but this incidence was similar for the oral semaglutide and placebo groups. In the insulin regimen analysis, most hypoglycaemic episodes generally occurred in patients on basal-bolus insulin. There were few on-treatment severe hypoglycaemic episodes, regardless of the background medication or insulin regimen.

There were few events of genital infection and increased urination in PIONEER 4 and 7, and no indication that the risk in patients treated with an SGLT2i was aggravated by concurrent treatment with oral semaglutide (see Table S4 in the “Supplementary Information”).

Discussion

In the primary analyses of the global PIONEER trials, oral semaglutide consistently demonstrated clinically meaningful reductions in HbA1c and body weight across a broad spectrum of patients with T2D [7,8,9,10,11,12,13,14]. In the current post-hoc exploratory subgroup analyses, the effects of oral semaglutide on HbA1c and body weight were broadly similar regardless of background medication and insulin regimen, and these effects were generally greater with oral semaglutide (14 mg and flexibly dosed) than with comparators. An exception was HbA1c reductions with liraglutide, which were similar to those seen with oral semaglutide. These observations are consistent with those from similar post-hoc analyses of patients receiving glucose-lowering background medication in the clinical trial programme for once-weekly subcutaneous semaglutide (SUSTAIN 2–4 and 10) [17].

In PIONEER 8, duration of diabetes, body weight and total daily insulin dose appeared to be higher in patients receiving basal-bolus insulin than in those on the other insulin regimens. This could be explained by patients requiring more comprehensive treatment as their disease progresses, therefore moving from a basal to a basal-bolus regimen. Patients receiving basal-bolus insulin had higher baseline body weight than other insulin regimens, which may relate to the fact that they were receiving higher doses and prandial doses of insulin, which are associated with weight gain. The treatment difference for change in HbA1c between oral semaglutide and placebo was smaller in patients taking insulin and metformin than in those taking insulin alone, although the treatment difference was still clinically relevant. It is unclear whether the efficacy of oral semaglutide or other GLP-1RAs is impacted by metformin in the presence of another insulin and other glucose-lowering background medications.

The current European Association for the Study of Diabetes (EASD) and American Diabetes Association (ADA) consensus report recommends stepwise addition of glucose-lowering medications, including the newer classes, i.e. GLP-1RAs, SGLT2is and dipeptidylpeptidase-4 inhibitors (DPP4is), with early combination therapy advocated when the patient’s HbA1c is above target despite diet, exercise and metformin [5, 6]. The preference for the medications to add after metformin depends on individual patient factors as well as other practical and economic considerations [5]. It is therefore important to understand the relative advantages and disadvantages of each treatment and how they might be used with other glucose-lowering agents in order to make informed decisions.

In these subgroup analyses, oral semaglutide 14 mg or flexibly dosed was more effective at reducing HbA1c and body weight than the DPP4i sitagliptin when added to a variety of representative background regimens. This is consistent with head-to-head clinical trials and meta-analyses which indicate that GLP-1RAs are more efficacious than DPP4is [18,19,20,21,22]. Furthermore, GLP-1RAs (and SGLT2is) lead to additional benefits that are not evident with DPP4is; for instance, they reduce cardiorenal risk [23,24,25,26,27,28,29,30].

There is considerable interest in the potential to combine GLP-1RAs with SGLT2is because their mechanisms of action are complementary with regard to target organs, disease processes [31], and the effects seen in cardiovascular/cardiorenal outcomes trials [23,24,25,26,27,28,29, 32,33,34,35,36]. There are relatively few prospective studies on the combined use of GLP-1RAs and SGLT2is, although a meta-analysis has suggested a benefit of the combination in terms of improved reductions in HbA1c and body weight [37]. In the current subgroup analysis of PIONEER 4, which included relatively few patients receiving both treatments, reductions in HbA1c (of 1.1–1.2%) and body weight (5 kg) were similar after 1 year in patients with T2D treated with oral semaglutide with or without background SGLT2i. The overall tolerability of the combination was similar to that of oral semaglutide alone.

In the PIONEER programme, oral semaglutide had a safety profile similar to that characterised for other GLP-1RAs [2]. In the current subgroup analyses, background medication had a limited effect on the tolerability of oral semaglutide (including on gastrointestinal AEs). The incidence of AEs tended to be higher for patients receiving insulin than other background medications, which may reflect the safety profile and the fact that insulin is often used later in the treatment course of T2D, when patients are more vulnerable. As might be expected, hypoglycaemic episodes were more common in patients receiving background insulin and sulphonylurea.

Efficacy and tolerability outcomes with oral semaglutide were generally consistent regardless of background insulin regimen in PIONEER 8, although AEs (including severe or blood glucose-confirmed hypoglycaemia) were more common with background basal insulin than with premixed or basal-bolus regimens. The total daily insulin dose requirement was reduced with oral semaglutide compared with placebo, without a clear link between insulin regimen subgroup and the change in total daily insulin dose. Given the increased risk of body weight gain and hypoglycaemia associated with long-term insulin use, the potential for reducing insulin doses with the addition of oral semaglutide (as well as its weight-decreasing potential) may be of clinical importance. In line with these results, in the randomised, placebo-controlled SUSTAIN 5 trial of once-weekly subcutaneous semaglutide as an add-on to basal insulin, the insulin dose was significantly decreased from baseline with both the semaglutide 0.5 and 1.0 mg doses compared with placebo [38].

Randomisation in these trials was prospectively stratified by baseline background medication, which can be considered a strength of the current analyses. Nevertheless, the subgroup analyses presented here were conducted post hoc and thus have limitations compared with prospective studies. Furthermore, the statistical analyses of efficacy outcomes were not controlled for multiplicity. As such, the outcomes should be considered exploratory in nature and warrant further prospective investigation. The PIONEER programme included a large number of patients, and data from more than 2800 were included in these analyses. Despite this, it should be noted that patient numbers in certain background medication subgroups were relatively low. Nevertheless, these analyses help to further characterise the profile of oral semaglutide in the context of guideline-recommended management of patients with T2D.

Conclusions

These exploratory subgroup analyses support the results of the primary analyses of the PIONEER trials and suggest that oral semaglutide can be used across a broad range of patients with T2D in combination with other commonly used glucose-lowering background medications.